In [1]:
pip install lime
Collecting lime
Downloading lime-0.2.0.1.tar.gz (275 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/275.7 kB ? eta -:--:--
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╸━ 266.2/275.7 kB 10.3 MB/s eta 0:00:01
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 275.7/275.7 kB 7.0 MB/s eta 0:00:00
Preparing metadata (setup.py) ... done
Requirement already satisfied: matplotlib in /usr/local/lib/python3.10/dist-packages (from lime) (3.8.0)
Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from lime) (1.26.4)
Requirement already satisfied: scipy in /usr/local/lib/python3.10/dist-packages (from lime) (1.13.1)
Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (from lime) (4.66.6)
Requirement already satisfied: scikit-learn>=0.18 in /usr/local/lib/python3.10/dist-packages (from lime) (1.5.2)
Requirement already satisfied: scikit-image>=0.12 in /usr/local/lib/python3.10/dist-packages (from lime) (0.24.0)
Requirement already satisfied: networkx>=2.8 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (3.4.2)
Requirement already satisfied: pillow>=9.1 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (11.0.0)
Requirement already satisfied: imageio>=2.33 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (2.36.1)
Requirement already satisfied: tifffile>=2022.8.12 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (2024.9.20)
Requirement already satisfied: packaging>=21 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (24.2)
Requirement already satisfied: lazy-loader>=0.4 in /usr/local/lib/python3.10/dist-packages (from scikit-image>=0.12->lime) (0.4)
Requirement already satisfied: joblib>=1.2.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn>=0.18->lime) (1.4.2)
Requirement already satisfied: threadpoolctl>=3.1.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn>=0.18->lime) (3.5.0)
Requirement already satisfied: contourpy>=1.0.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (1.3.1)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (0.12.1)
Requirement already satisfied: fonttools>=4.22.0 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (4.55.1)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (1.4.7)
Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (3.2.0)
Requirement already satisfied: python-dateutil>=2.7 in /usr/local/lib/python3.10/dist-packages (from matplotlib->lime) (2.8.2)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.7->matplotlib->lime) (1.16.0)
Building wheels for collected packages: lime
Building wheel for lime (setup.py) ... done
Created wheel for lime: filename=lime-0.2.0.1-py3-none-any.whl size=283834 sha256=203b19bbee9cdc8a60477d5e53e57fda6388eb2f3420e1b2c133fa03d0efe1d9
Stored in directory: /root/.cache/pip/wheels/fd/a2/af/9ac0a1a85a27f314a06b39e1f492bee1547d52549a4606ed89
Successfully built lime
Installing collected packages: lime
Successfully installed lime-0.2.0.1
In [2]:
# Importing necessary packages
# for segmentation part
import os
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import glob as gb
import tensorflow as tf
from sklearn.model_selection import train_test_split
from tensorflow.keras import layers, Model
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, BatchNormalization, Conv2DTranspose, Activation, add
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.losses import BinaryCrossentropy
from tensorflow.keras.metrics import Precision, Recall, MeanSquaredError
from tensorflow.keras import backend as K
sns.set_theme(style='darkgrid', palette='pastel')
#from tensorflow.keras.models import load_model
import keras
#Occlusion
from PIL import Image
import copy
import math
from keras.models import load_model
# for classification part
import os
import numpy as np
import pandas as pd
import random
import matplotlib.pyplot as plt
import seaborn as sns
from PIL import Image
import torch
import torch.onnx
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms, models
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score, accuracy_score
from tqdm import tqdm
from torchvision.transforms import functional as F
from lime import lime_image
from skimage.segmentation import mark_boundaries
from numpy.random import default_rng
from PIL import Image
import os
import matplotlib.pyplot as plt
from PIL import Image
from torchvision import transforms, models
from sklearn.metrics import classification_report, confusion_matrix, roc_auc_score, accuracy_score
from tqdm import tqdm
import zipfile
from google.colab import drive
In [3]:
# Set seed for reproducibility
fixed_seed = 42
random.seed(fixed_seed)
np.random.seed(fixed_seed)
torch.manual_seed(fixed_seed)
torch.cuda.manual_seed_all(fixed_seed)
# Ensure deterministic behavior for PyTorch
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
Link for the google drive folder, where there are the trained models and the dataset: https://drive.google.com/drive/folders/1q_ifXgKSiG2WbrCW2mn65fJZsdsT0HpG?usp=sharing
1 Classification: MobileVNet3¶
In [4]:
# You have to set it to your drive link later on, all that paths.
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
In [5]:
# Unzip the downloaded file
with zipfile.ZipFile("/content/drive/MyDrive/XAI/SkinCancer.zip", 'r') as zip_ref:
zip_ref.extractall("extracted_folder")
# List contents
import os
os.listdir("extracted_folder")
Out[5]:
['images', 'GroundTruth.csv', 'masks']
In [6]:
# Check if CUDA is available and set the device accordingly
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")
# Data Preparation
columns_rearranged = ['image', 'NV', 'MEL', 'BCC', 'AKIEC', 'DF', 'VASC']
#skin_cancer_df = pd.read_csv('/content/drive/My Drive/AI_and_Vis/A2_data_from_Kaggle/GroundTruth.csv')
skin_cancer_df = pd.read_csv('/content/extracted_folder/GroundTruth.csv')
skin_cancer_df.drop(columns=['BKL'], inplace=True)
skin_cancer_df = skin_cancer_df[columns_rearranged]
multiclass_categories = {
0: 'NV',
1: 'MEL',
2: 'BCC',
3: 'AKIEC',
4: 'DF',
5: 'VASC'
}
filenames_list, multiclass_categories_list = [], []
base_paths = [
("/content/extracted_folder/images", skin_cancer_df),
]
"""
base_paths = [
("images", skin_cancer_df),
]
"""
Using device: cuda
Out[6]:
'\nbase_paths = [\n ("images", skin_cancer_df),\n]\n'
In [7]:
for base_path, database in base_paths:
if isinstance(database, pd.DataFrame):
for row in database.itertuples(index=False):
filename = os.path.join(base_path, row[0]) + '.jpg'
filenames_list.append(filename)
multiclass_index = np.argmax(row[1:])
multiclass_categories_list.append(multiclass_index)
else:
for sub_directory in database[0]:
for sub_sub_directory in database[1]:
final_path = os.path.join(base_path, sub_directory, sub_sub_directory)
for filename in os.listdir(final_path):
filenames_list.append(os.path.join(final_path, filename))
if sub_sub_directory in ['nevus', 'NotMelanoma']:
multiclass_categories_list.append(0)
else:
multiclass_categories_list.append(database[1].index(sub_sub_directory))
multiclass_df = pd.DataFrame({"filename": filenames_list, "category": multiclass_categories_list})
multiclass_df = multiclass_df.sample(frac=1).reset_index(drop=True)
multiclass_class_counts = np.bincount(multiclass_categories_list)
multiclass_class_weights = {i: len(multiclass_categories_list) / (2.0 * count) for i, count in enumerate(multiclass_class_counts)}
print(f"Multi-class weights: {multiclass_class_weights}")
Multi-class weights: {0: 0.6416581240389544, 1: 4.499101527403414, 2: 9.742217898832685, 3: 15.313455657492355, 4: 43.54347826086956, 5: 35.264084507042256}
In [8]:
# Visualization
label, count = np.unique(multiclass_df.category, return_counts=True)
uni = pd.DataFrame(data=count, index=multiclass_categories.values(), columns=['Count'])
plt.figure(figsize=(14, 4), dpi=200)
sns.barplot(data=uni, x=uni.index, y='Count', palette='icefire', width=0.4).set_title('Class distribution in Dataset', fontsize=15)
plt.show()
<ipython-input-8-a76bd2b1a029>:6: FutureWarning:
Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.
sns.barplot(data=uni, x=uni.index, y='Count', palette='icefire', width=0.4).set_title('Class distribution in Dataset', fontsize=15)
As we can see, the original dataset is heavily imbalanced towards the class “NV”, considerably less amount of samples are present in the dataset from other classes, especially from classes “DF” and “VASC”. This plot can give useful information for model builders & developers when constructing or improving a machine learning model on this skin cancer dataset.
1.1 DataLoading for classification¶
In [9]:
dataset_path = "/content/extracted_folder"
# dataset_path = "images"
valid_images = []
for img_file in os.listdir(dataset_path):
try:
img_path = os.path.join(dataset_path, img_file)
with Image.open(img_path) as img:
img.verify() # Verify that the file is intact
valid_images.append(img_file)
except Exception as e:
print(f"Skipping corrupted image: {img_file}, Error: {e}")
Skipping corrupted image: images, Error: [Errno 21] Is a directory: '/content/extracted_folder/images' Skipping corrupted image: GroundTruth.csv, Error: cannot identify image file '/content/extracted_folder/GroundTruth.csv' Skipping corrupted image: masks, Error: [Errno 21] Is a directory: '/content/extracted_folder/masks'
In [10]:
# Train-validation-test split
multiclass_train_df, multiclass_temp_df = train_test_split(multiclass_df, test_size=0.2, stratify=multiclass_df['category'], random_state=fixed_seed)
multiclass_cv_df, multiclass_test_df = train_test_split(multiclass_temp_df, test_size=0.5, stratify=multiclass_temp_df['category'], random_state=fixed_seed)
class SkinCancerDataset(Dataset):
def __init__(self, dataframe, transform=None):
self.dataframe = dataframe
self.transform = transform
def __len__(self):
return len(self.dataframe)
def __getitem__(self, idx):
img_path = self.dataframe.iloc[idx, 0]
image = Image.open(img_path).convert('RGB')
label = int(self.dataframe.iloc[idx, 1])
if self.transform:
image = self.transform(image)
return image, label
class DeterministicTransform:
def __init__(self, seed):
self.seed = seed
def __call__(self, img):
random.seed(self.seed) # Set the random seed for consistency
np.random.seed(self.seed)
# Define deterministic augmentations
if random.random() < 0.5: # Horizontal Flip
img = transforms.functional.hflip(img)
if random.random() < 0.5: # Vertical Flip
img = transforms.functional.vflip(img)
angle = random.uniform(-20, 20) # Rotation
img = transforms.functional.rotate(img, angle)
return img
image_size = (224, 224)
batch_size = 32
# transform = transforms.Compose([
# transforms.Resize(image_size),
# transforms.RandomHorizontalFlip(),
# transforms.RandomVerticalFlip(),
# transforms.RandomRotation(20),
# transforms.ToTensor(),
# transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
# ])
transform = transforms.Compose([
transforms.Resize(image_size),
DeterministicTransform(seed=fixed_seed),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
train_dataset = SkinCancerDataset(multiclass_train_df, transform=transform)
val_dataset = SkinCancerDataset(multiclass_cv_df, transform=transform)
test_dataset = SkinCancerDataset(multiclass_test_df, transform=transform)
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=14)
val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=14)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, num_workers=14)
/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py:617: UserWarning: This DataLoader will create 14 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary. warnings.warn(
1.2 Classification: MobileNetV3¶
Source for the model implementation: https://www.kaggle.com/code/dariusfoodeei/multiclass-skincancer-torch
In [11]:
class MobileNetV3(nn.Module):
def __init__(self, num_classes):
super(MobileNetV3, self).__init__()
self.model = models.mobilenet_v3_small(pretrained=True)
in_features = self.model.classifier[3].in_features
self.model.classifier[3] = nn.Linear(in_features, num_classes)
self.softmax = nn.Softmax(dim=1)
def forward(self, x):
x = self.model(x)
x = self.softmax(x)
return x
model_class = MobileNetV3(num_classes=len(multiclass_categories))
model_class = model_class.to(device)
[torch.cuda.get_device_name(i) for i in range(torch.cuda.device_count())]
model_class = MobileNetV3(num_classes=len(multiclass_categories))
model_class = model_class.to(device)
# Define the criterion with class weights
class_weights = torch.tensor(list(multiclass_class_weights.values()), dtype=torch.float32).to(device)
criterion = nn.CrossEntropyLoss(weight=class_weights).to(device)
optimizer = optim.Adam(model_class.parameters(), lr=3e-4)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'min', factor=0.5, patience=2)
num_epochs = 20
early_stopping_patience = 6
best_val_loss = float('inf')
epochs_no_improve = 0
# This might result in partially loaded images, so proceed with caution !!!!
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
def train_one_epoch(epoch, model, train_loader, optimizer, criterion, device):
model.train()
train_loss = 0.0
correct_train = 0
for images, labels in tqdm(train_loader, desc=f"Training Epoch {epoch+1}/{num_epochs}", ncols=100):
images, labels = images.to(device, dtype=torch.float32), labels.cuda().long()
optimizer.zero_grad()
outputs = model(images)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
train_loss += loss.item() * images.size(0)
_, predicted = torch.max(outputs, 1)
correct_train += (predicted == labels).sum().item()
train_loss /= len(train_loader.dataset)
train_accuracy = correct_train / len(train_loader.dataset)
return train_loss, train_accuracy
def validate_one_epoch(epoch, model, val_loader, criterion, device):
model.eval()
val_loss = 0.0
correct_val = 0
with torch.no_grad():
for images, labels in tqdm(val_loader, desc=f"Validation Epoch {epoch+1}/{num_epochs}", ncols=100):
images, labels = images.to(device, dtype=torch.float32), labels.to(device, dtype=torch.long)
outputs = model(images)
loss = criterion(outputs, labels)
val_loss += loss.item() * images.size(0)
_, predicted = torch.max(outputs, 1)
correct_val += (predicted == labels).sum().item()
val_loss /= len(val_loader.dataset)
val_accuracy = correct_val / len(val_loader.dataset)
return val_loss, val_accuracy
def train_model(model, train_loader, val_loader, optimizer, scheduler, num_epochs, early_stopping_patience, device):
best_val_loss = float('inf')
epochs_no_improve = 0
for epoch in range(num_epochs):
train_loss, train_accuracy = train_one_epoch(epoch, model, train_loader, optimizer, criterion, device)
val_loss, val_accuracy = validate_one_epoch(epoch, model, val_loader, criterion, device)
print(f"Epoch {epoch+1}/{num_epochs}, Train Loss: {train_loss:.4f}, Train Accuracy: {train_accuracy:.4f}, "
f"Validation Loss: {val_loss:.4f}, Validation Accuracy: {val_accuracy:.4f}")
scheduler.step(val_loss)
if val_loss < best_val_loss:
best_val_loss = val_loss
torch.save(model.state_dict(), 'best_model.pth')
if epochs_no_improve >= early_stopping_patience:
print("Early stopping")
break
# This is commented out to not to train the model again
# train_model(model_class, train_loader, val_loader, optimizer, scheduler, num_epochs, early_stopping_patience, device)
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead. warnings.warn( /usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=MobileNet_V3_Small_Weights.IMAGENET1K_V1`. You can also use `weights=MobileNet_V3_Small_Weights.DEFAULT` to get the most up-to-date weights. warnings.warn(msg) Downloading: "https://download.pytorch.org/models/mobilenet_v3_small-047dcff4.pth" to /root/.cache/torch/hub/checkpoints/mobilenet_v3_small-047dcff4.pth 100%|██████████| 9.83M/9.83M [00:00<00:00, 59.0MB/s]
Load best classification model
In [12]:
# Load the best model (PyTorch format)
model_class = MobileNetV3(num_classes=len(multiclass_categories))
model_class.load_state_dict(torch.load('/content/drive/MyDrive/XAI/best_model_classification.pth', map_location=torch.device('cuda' if torch.cuda.is_available() else 'cpu')))
model_class.eval()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model_class.to(device)
<ipython-input-12-7b8ad7cc3b66>:4: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
model_class.load_state_dict(torch.load('/content/drive/MyDrive/XAI/best_model_classification.pth', map_location=torch.device('cuda' if torch.cuda.is_available() else 'cpu')))
Out[12]:
MobileNetV3(
(model): MobileNetV3(
(features): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(16, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=16, bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(16, 8, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(8, 16, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(2): Conv2dNormActivation(
(0): Conv2d(16, 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(2): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(16, 72, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(72, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Conv2dNormActivation(
(0): Conv2d(72, 72, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=72, bias=False)
(1): BatchNorm2d(72, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(2): Conv2dNormActivation(
(0): Conv2d(72, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(24, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(3): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(24, 88, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(88, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Conv2dNormActivation(
(0): Conv2d(88, 88, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=88, bias=False)
(1): BatchNorm2d(88, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(2): Conv2dNormActivation(
(0): Conv2d(88, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(24, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(4): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(24, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96, bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(96, 24, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(24, 96, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(96, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(5): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(240, 64, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(64, 240, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(6): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(240, 64, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(64, 240, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(7): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 120, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(120, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
(1): BatchNorm2d(120, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(120, 32, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(32, 120, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(120, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(48, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(8): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(48, 144, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(144, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(144, 144, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=144, bias=False)
(1): BatchNorm2d(144, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(144, 40, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(40, 144, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(144, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(48, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(9): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(48, 288, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(288, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(288, 288, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=288, bias=False)
(1): BatchNorm2d(288, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(288, 72, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(72, 288, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(288, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(10): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(576, 576, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=576, bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(576, 144, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(144, 576, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(576, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(11): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(576, 576, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=576, bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(576, 144, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(144, 576, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(576, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(12): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
)
(avgpool): AdaptiveAvgPool2d(output_size=1)
(classifier): Sequential(
(0): Linear(in_features=576, out_features=1024, bias=True)
(1): Hardswish()
(2): Dropout(p=0.2, inplace=True)
(3): Linear(in_features=1024, out_features=6, bias=True)
)
)
(softmax): Softmax(dim=1)
)
In [13]:
# Test the model
model_class.eval()
test_loss = 0.0
correct_test = 0
# Create a dummy input tensor matching the input dimensions of the model
dummy_input = torch.randn(1, 3, 224, 224, device='cuda')
# Export the model
#torch.onnx.export(model, dummy_input, "model.onnx", export_params=True, opset_version=10)
predictions = []
true_classes = []
with torch.no_grad():
for images, labels in tqdm(test_loader, desc="Testing", ncols=100):
images, labels = images.to(device), labels.to(device)
outputs = model_class(images)
loss = criterion(outputs, labels)
test_loss += loss.item() * images.size(0)
_, predicted = torch.max(outputs, 1)
correct_test += (predicted == labels).sum().item()
predictions.extend(predicted.cpu().numpy())
true_classes.extend(labels.cpu().numpy())
test_loss = test_loss / len(test_loader.dataset)
test_accuracy = correct_test / len(test_loader.dataset)
print(f'Test Loss: {test_loss:.4f}')
print(f'Test Accuracy: {test_accuracy*100:.2f}%')
# Calculate precision
precision = accuracy_score(true_classes, predictions) * 100
print(f'Precision: {precision:.2f}%')
Testing: 100%|██████████████████████████████████████████████████████| 32/32 [00:19<00:00, 1.62it/s]
Test Loss: 1.3591 Test Accuracy: 76.05% Precision: 76.05%
In [14]:
# Confusion matrix
conf_matrix = confusion_matrix(true_classes, predictions)
plt.figure(figsize=(8, 6), dpi=200)
sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='icefire', xticklabels=multiclass_categories.keys(), yticklabels=multiclass_categories.keys())
plt.xlabel('Predicted')
plt.ylabel('Actual')
plt.title('Confusion Matrix')
plt.show()
This plot shows the confusion matrix on the test set produced by the best trained classification model. The confusion matrix also underlines our previous findings about the dataset, namely that it is heavily imbalanced and most instances belong to class “NV” (0). As we can see, the model performs well on the majority class „NV” (0) (637 correct predictions), but seems to struggle for class „MEL” (1) (41 correct predictions versus 68 misclassifications into class 0). It also underperforms for class „DF” (4) and „VASC”(5) likely due to the insufficient representation of these classes in the dataset. This confusion matrix can be insightful for model builders & developers when evaluating and improving their trained machine learning model.
2 Segmentation: Fuzzy U-net¶
Source for this model: https://www.kaggle.com/code/iakhtar0/2-skin-cancer-images-segmentation-fuzzy-unet
2.1 Segmentation plots¶
In [15]:
# images and masks folder paths
image_dir = '/content/extracted_folder/images'
mask_dir = '/content/extracted_folder/masks'
# list of images and masks from the directory
image_files = sorted([f for f in os.listdir(image_dir) if f.endswith('.jpg')])
mask_files = sorted([f for f in os.listdir(mask_dir) if f.endswith('.png')])
# generating lis of image names (without extension) & masks names (without 'segmentation' suffix & extension)
image_names = [os.path.splitext(f)[0] for f in image_files]
mask_names = [os.path.splitext(f)[0].replace('_segmentation', '') for f in mask_files]
# checking if all the images have corresponding masks or not
missing_masks = [img for img in image_names if img not in mask_names]
if len(missing_masks) == 0:
print('All images have corresponding masks...')
else:
print('There are some images that dont have masks...')
All images have corresponding masks...
In [16]:
# function to display image & mask
def display_function(image_path, mask_path):
image = Image.open(image_path)
mask = Image.open(mask_path)
# side by side display
fig, axis = plt.subplots(1, 2, figsize=(10, 5))
axis[0].imshow(image)
axis[0].set_title('Skin Image')
axis[0].axis('off')
axis[1].imshow(mask, cmap='gray')
axis[1].set_title('Mask')
axis[1].axis('off')
plt.show()
# Displaying 5 images & masks
for i in range(5):
image_path = os.path.join(image_dir, image_files[i])
mask_name = os.path.splitext(image_files[i])[0] + '_segmentation.png'
mask_path = os.path.join(mask_dir, mask_name)
display_function(image_path, mask_path)
In [17]:
# function to display masked images
def display_masked_images(image_path, mask_path):
image = Image.open(image_path)
mask = Image.open(mask_path)
# converting image & mask to numpy array
image_np = np.array(image)
mask_np = np.array(mask)
return image_np, mask_np
# displaying 5 masked images
fig, axis = plt.subplots(1, 5, figsize=(20,5))
for i in range(5):
image_path = os.path.join(image_dir, image_files[i])
mask_name = os.path.splitext(image_files[i])[0] + '_segmentation.png'
mask_path = os.path.join(mask_dir, mask_name)
image_np, mask_np = display_masked_images(image_path, mask_path)
# display masked images
axis[i].imshow(image_np)
axis[i].imshow(mask_np, cmap='Reds', alpha=0.5) # alpha - to make it semi transparent
axis[i].axis('off')
plt.show()
2.2 DataLoading for segmentation¶
In [18]:
# setting the default values
IMG_SIZE = 256
BATCH_SIZE = 32
BUFFER_SIZE = 1000
AUTOTUNE = tf.data.experimental.AUTOTUNE
IMG_DIR = '/content/extracted_folder/images'
MASK_DIR = '/content/extracted_folder/masks'
# function to get image path and masks
def image_mask_paths(image_dir, mask_dir):
images_path = sorted(gb.glob(os.path.join(image_dir, '*.jpg')))
masks_path = sorted(gb.glob(os.path.join(mask_dir, '*.png')))
images_path = np.array(images_path)
masks_path = np.array(masks_path)
return images_path, masks_path
images_path, masks_path = image_mask_paths(IMG_DIR, MASK_DIR)
print("Number of images:", len(images_path))
print("Number of masks:", len(masks_path))
Number of images: 10015 Number of masks: 10015
In [19]:
X_train, X_temp, y_train, y_temp = train_test_split(images_path, masks_path, train_size=0.8, random_state=fixed_seed)
X_val, X_test, y_val, y_test = train_test_split(X_temp, y_temp, train_size=0.5, random_state=fixed_seed)
print(f'Training Samples: {len(X_train)}')
print(f'Validation Samples: {len(X_val)}')
print(f'Testing Samples: {len(X_test)}')
# function to loading & preprocessing data
def mapping_function(image_path, mask_path):
img = tf.io.read_file(image_path)
img = tf.image.decode_jpeg(img, channels=3)
img = tf.image.resize(img, (IMG_SIZE, IMG_SIZE))
img = tf.cast(img, tf.float32)/255.0
mask = tf.io.read_file(mask_path)
mask = tf.image.decode_jpeg(mask, channels=1)
mask = tf.image.resize(mask, (IMG_SIZE, IMG_SIZE))
mask = tf.cast(mask, tf.float32) / 255.0
return img, mask
# creating training set
train_set = tf.data.Dataset.from_tensor_slices((X_train, y_train))
train_set = train_set.map(mapping_function, num_parallel_calls=AUTOTUNE)
train_set = train_set.shuffle(BUFFER_SIZE).batch(BATCH_SIZE).prefetch(buffer_size=AUTOTUNE)
# creating validation set
validation_set = tf.data.Dataset.from_tensor_slices((X_val, y_val))
validation_set = validation_set.map(mapping_function, num_parallel_calls=AUTOTUNE)
validation_set = validation_set.shuffle(BUFFER_SIZE).batch(BATCH_SIZE).prefetch(buffer_size=AUTOTUNE)
# creating testing set
test_set = tf.data.Dataset.from_tensor_slices((X_test, y_test))
test_set = test_set.map(mapping_function, num_parallel_calls=AUTOTUNE)
test_set = test_set.shuffle(BUFFER_SIZE).batch(BATCH_SIZE).prefetch(buffer_size=AUTOTUNE)
Training Samples: 8012 Validation Samples: 1001 Testing Samples: 1002
2.3 Segmentation: Fuzzy U-net implementation¶
In [20]:
def fuzzy_unet(input_shape=(IMG_SIZE, IMG_SIZE, 3)):
inputs = Input(shape=input_shape)
x = Conv2D(32, 3, padding="same")(inputs)
x = BatchNormalization()(x)
x = Activation("relu")(x)
skip_connections = []
# Encoder
for filters in [64, 128, 256, 512]:
x = Conv2D(filters, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
x = Conv2D(filters, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
skip_connections.append(x)
x = MaxPooling2D(2, padding="same")(x)
# Bottleneck with fuzzy membership calculations
x = Conv2D(1024, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
x = Conv2D(1024, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
# Decoder with fuzzy skip connections
for filters in [512, 256, 128, 64]:
x = Conv2DTranspose(filters, 2, strides=2, padding="same")(x)
# Fuzzy attention for skip connections
skip_connection = skip_connections.pop()
attention = Conv2D(filters, 1, padding='same', activation='sigmoid')(skip_connection)
skip_connection = skip_connection * attention # Fuzzy weighting
x = add([x, skip_connection])
x = Conv2D(filters, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
x = Conv2D(filters, 3, padding='same')(x)
x = BatchNormalization()(x)
x = Activation("relu")(x)
outputs = Conv2D(1, 1, activation="sigmoid")(x)
model = Model(inputs, outputs)
return model
# Instantiate the fuzzy U-Net model
model_seg = fuzzy_unet(input_shape=(IMG_SIZE, IMG_SIZE, 3))
# model.summary()
Loss functions
In [21]:
# Dice Coefficient
def dice_coefficient(y_true, y_pred, smooth=1):
intersection = tf.reduce_sum(y_true * y_pred, axis=[1, 2, 3])
union = tf.reduce_sum(y_true, axis=[1, 2, 3]) + tf.reduce_sum(y_pred, axis=[1, 2, 3])
dice = (2 * intersection + smooth) / (union + smooth)
return tf.reduce_mean(dice, axis=0)
# Dice Loss
def dice_coef_loss(y_true, y_pred):
return 1 - dice_coefficient(y_true, y_pred)
# IoU Metric
def iou(y_true, y_pred, smooth=1):
intersection = tf.reduce_sum(y_true * y_pred, axis=[1, 2, 3])
union = tf.reduce_sum(y_true, axis=[1, 2, 3]) + tf.reduce_sum(y_pred, axis=[1, 2, 3]) - intersection
iou_score = (intersection + smooth) / (union + smooth)
return tf.reduce_mean(iou_score, axis=0)
In [22]:
model_seg.compile(
optimizer=Adam(learning_rate=0.0002),
loss=BinaryCrossentropy(),
metrics=[
'binary_accuracy',
Precision(name='precision'),
Recall(name='recall'),
dice_coefficient,
iou,
MeanSquaredError(name='mse')
]
)
Loading the trained model with its weights:
In [23]:
model_seg = load_model("/content/drive/MyDrive/XAI/Skin_cancer_Segmentation.keras", custom_objects={"dice_coefficient": dice_coefficient, "iou": iou})
3 XAI approaches¶
3.1 XAI approaches for the classification model¶
3.1.1. Visualize class prediction probabilities¶
In [24]:
# Define function to visualize classified images with all class scores
def plot_classified_images_with_scores(model, dataloader, class_names, device, num_images=6):
"""
Plot correctly and incorrectly classified images with their probabilities for all classes.
Args:
- model: Trained PyTorch model.
- dataloader: DataLoader for the test dataset.
- class_names: List of class names.
- device: Computation device ('cuda' or 'cpu').
- num_images: Number of images to plot (default: 6).
"""
model.eval()
all_images = []
all_labels = []
all_predictions = []
all_probs = []
# Collect predictions, true labels, and probabilities
with torch.no_grad():
for images, labels in dataloader:
images, labels = images.to(device), labels.to(device)
outputs = model(images)
probs = torch.softmax(outputs, dim=1)
_, preds = torch.max(outputs, 1)
all_images.append(images.cpu())
all_labels.append(labels.cpu())
all_predictions.append(preds.cpu())
all_probs.append(probs.cpu())
# Flatten lists
all_images = torch.cat(all_images, dim=0)
all_labels = torch.cat(all_labels, dim=0)
all_predictions = torch.cat(all_predictions, dim=0)
all_probs = torch.cat(all_probs, dim=0)
# Identify correct and incorrect indices
correct_indices = (all_predictions == all_labels).nonzero(as_tuple=True)[0]
incorrect_indices = (all_predictions != all_labels).nonzero(as_tuple=True)[0]
# # Select random correct and incorrect samples
# # Generate choices
# correct_sample_indices = np.random.choice(correct_indices, num_images // 2, replace=False)
# incorrect_sample_indices = np.random.choice(incorrect_indices, num_images // 2, replace=False)
# selected_indices = np.concatenate([correct_sample_indices, incorrect_sample_indices])
# # Sort indices for reproducibility
# correct_indices = torch.sort(correct_indices)[0]
# incorrect_indices = torch.sort(incorrect_indices)[0]
# # Select random correct and incorrect samples
# np.random.seed(fixed_seed) # Ensure reproducibility for the choice
# correct_sample_indices = np.random.choice(correct_indices, num_images // 2, replace=False)
# incorrect_sample_indices = np.random.choice(incorrect_indices, num_images // 2, replace=False)
# selected_indices = np.concatenate([correct_sample_indices, incorrect_sample_indices])
# Sort indices for reproducibility
correct_indices = torch.sort(correct_indices)[0]
incorrect_indices = torch.sort(incorrect_indices)[0]
# Ensure consistent image selection
num_correct_samples = min(len(correct_indices), num_images // 2)
num_incorrect_samples = min(len(incorrect_indices), num_images // 2)
correct_sample_indices = correct_indices[:num_correct_samples]
incorrect_sample_indices = incorrect_indices[:num_incorrect_samples]
selected_indices = torch.cat([correct_sample_indices, incorrect_sample_indices])
# Plot the images
fig, axes = plt.subplots(2, num_images, figsize=(20, 10)) # Two rows: images and bar plots
for i, idx in enumerate(selected_indices):
image = all_images[idx].permute(1, 2, 0).numpy() # Convert to HWC format
image = np.clip(image * [0.229, 0.224, 0.225] + [0.485, 0.456, 0.406], 0, 1) # Denormalize
true_label = class_names[all_labels[idx]]
pred_label = class_names[all_predictions[idx]]
pred_probs = all_probs[idx].numpy()
# Image Plot
ax_image = axes[0, i]
ax_image.imshow(image)
ax_image.axis('off')
ax_image.set_title(
f"Pred: {pred_label}\nTrue: {true_label}",
color=("green" if all_predictions[idx] == all_labels[idx] else "red"),
fontsize=10
)
# Bar Plot for Probabilities
ax_bar = axes[1, i]
ax_bar.bar(class_names, pred_probs, color='blue', alpha=0.7)
ax_bar.set_ylim([0, 1])
ax_bar.set_xticks(range(len(class_names)))
ax_bar.set_xticklabels(class_names, rotation=45, fontsize=8)
ax_bar.set_title("Class Probabilities", fontsize=10)
plt.tight_layout()
plt.show()
class_names = list(multiclass_categories.values())
plot_classified_images_with_scores(model_class, test_loader, class_names, device)
/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py:617: UserWarning: This DataLoader will create 14 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary. warnings.warn(
This plot shows 3 correctly (green text) and 3 incorrectly (red text) classified instances from the test set, as well as the class prediction probabilities corresponding to the classes. As we can see from the bar plots, the classification model usually assigns the highest probability (4 out of 6 cases) to class NV. (As we have mentioned before, the original dataset is heavily imbalanced towards the class “NV”.) However, even for this class the model is not so confident, as it assigns less than 0.4 probability to this class in all cases, and also 2 out of the 3 misclassified examples were “NV” samples misclassified as either a “MEL” or an “AKIEC” sample. Overall, we can see that the model is not so certain about its predictions, and this can be a valuable insight for both model builders & developers who would like to see how robust the classification model is, how uncertain it is about its predictions, and for non-experts (clinicians, doctors, or laypersons who are the users of different skin cancer detection apps) to properly calibrate their trust towards the skin cancer classification model.
3.1.2. Grad-Cam for the classification model¶
In [25]:
def preprocess_image(img_path, transform):
"""Load and preprocess an image for the model."""
image = Image.open(img_path).convert("RGB")
return transform(image).unsqueeze(0)
def generate_gradcam_heatmap(model, img_tensor, target_class, last_conv_layer_name):
"""Generate Grad-CAM heatmap."""
# Ensure model is in evaluation mode
model.eval()
# Hook to capture gradients
gradients = []
activations = []
def backward_hook(module, grad_in, grad_out):
gradients.append(grad_out[0])
def forward_hook(module, input, output):
activations.append(output)
# Register hooks to the specified layer
last_conv_layer = dict(model.named_modules())[last_conv_layer_name]
last_conv_layer.register_forward_hook(forward_hook)
last_conv_layer.register_backward_hook(backward_hook)
# Forward pass
img_tensor = img_tensor.to(device)
outputs = model(img_tensor)
target_score = outputs[0, target_class]
# Backward pass to compute gradients of the target class
model.zero_grad()
target_score.backward(retain_graph=True)
# Extract gradients and activations
grads = gradients[0].cpu().data.numpy()
acts = activations[0].cpu().data.numpy()
# Compute importance weights
weights = np.mean(grads, axis=(2, 3)) # Global average pooling
heatmap = np.sum(weights[:, :, np.newaxis, np.newaxis] * acts, axis=1).squeeze()
# Normalize the heatmap
heatmap = np.maximum(heatmap, 0)
heatmap /= heatmap.max() if heatmap.max() != 0 else 1
return heatmap
def overlay_gradcam_heatmap(heatmap, original_image, alpha=0.6, colormap='jet'):
"""Overlay Grad-CAM heatmap on the original image."""
heatmap = np.uint8(255 * heatmap)
colormap = plt.get_cmap(colormap)
colored_heatmap = colormap(heatmap)
colored_heatmap = np.delete(colored_heatmap, 3, axis=2) # Remove alpha channel
# Resize heatmap to original image dimensions
colored_heatmap = Image.fromarray((colored_heatmap * 255).astype("uint8")).resize(original_image.size, Image.LANCZOS)
colored_heatmap = np.array(colored_heatmap) / 255.0 # Normalize to [0, 1]
# Superimpose the heatmap onto the original image
original_image = np.array(original_image) / 255.0
superimposed_image = colored_heatmap * alpha + original_image
superimposed_image = np.clip(superimposed_image, 0, 1)
return (superimposed_image * 255).astype("uint8")
# EASY IMAGE
# Load and preprocess the image
#image_path = "/content/drive/My Drive/AI_and_Vis/A2_data_from_Kaggle/extracted_images_and_masks/images/ISIC_0024317.jpg" # ISIC_0024624.jpg
image_path = "/content/extracted_folder/images/ISIC_0024810.jpg" # ISIC_0024624.jpg
original_image = Image.open(image_path).convert("RGB")
img_tensor = preprocess_image(image_path, transform)
# Generate Grad-CAM heatmap
target_class = 0 # Update to the desired class index
last_conv_layer_name = "model.features.11.block"
heatmap = generate_gradcam_heatmap(model_class, img_tensor, target_class, last_conv_layer_name)
# Overlay Grad-CAM heatmap on the original image
superimposed_image = overlay_gradcam_heatmap(heatmap, original_image, alpha=0.6)
# Display the results
plt.figure(figsize=(14, 7))
# Original Image
plt.subplot(1, 3, 1)
plt.imshow(original_image)
plt.title("Original Image")
plt.axis("off")
# Grad-CAM Overlay
plt.subplot(1, 3, 2)
plt.imshow(superimposed_image)
plt.title(f"Grad-CAM Overlay (Class {target_class})")
plt.axis("off")
# Heatmap with Colorbar
plt.subplot(1, 3, 3)
plt.imshow(heatmap, cmap="jet")
plt.title("Heatmap (Color Scale)")
plt.axis("off")
cbar = plt.colorbar(plt.cm.ScalarMappable(cmap="jet"), ax=plt.gca(), fraction=0.046, pad=0.04)
cbar.set_label("Activation Intensity")
plt.tight_layout()
plt.show()
# CHALLENGING IMAGE
#image_path_2 = "/content/drive/My Drive/AI_and_Vis/A2_data_from_Kaggle/extracted_images_and_masks/images/ISIC_0024624.jpg"
image_path_2 = "/content/extracted_folder/images/ISIC_0024354.jpg"
original_image = Image.open(image_path_2).convert("RGB")
img_tensor = preprocess_image(image_path_2, transform)
# Generate Grad-CAM heatmap
target_class = 0 # Update to the desired class index
last_conv_layer_name = "model.features.11.block"
heatmap = generate_gradcam_heatmap(model_class, img_tensor, target_class, last_conv_layer_name)
# Overlay Grad-CAM heatmap on the original image
superimposed_image = overlay_gradcam_heatmap(heatmap, original_image, alpha=0.6)
# Display the results
plt.figure(figsize=(14, 7))
# Original Image
plt.subplot(1, 3, 1)
plt.imshow(original_image)
plt.title("Original Image")
plt.axis("off")
# Grad-CAM Overlay
plt.subplot(1, 3, 2)
plt.imshow(superimposed_image)
plt.title(f"Grad-CAM Overlay (Class {target_class})")
plt.axis("off")
# Heatmap with Colorbar
plt.subplot(1, 3, 3)
plt.imshow(heatmap, cmap="jet")
plt.title("Heatmap (Color Scale)")
plt.axis("off")
cbar = plt.colorbar(plt.cm.ScalarMappable(cmap="jet"), ax=plt.gca(), fraction=0.046, pad=0.04)
cbar.set_label("Activation Intensity")
plt.tight_layout()
plt.show()
/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py:1827: FutureWarning: Using a non-full backward hook when the forward contains multiple autograd Nodes is deprecated and will be removed in future versions. This hook will be missing some grad_input. Please use register_full_backward_hook to get the documented behavior. self._maybe_warn_non_full_backward_hook(args, result, grad_fn)
These Grad-CAM plots visualize the regions of input images (above input image with sharp bordered lesion, below input image with blurry bordered lesion) that contributed most to the model's prediction for class „NV” (0). Moreover, the accompanying heatmaps provide an even clearer activation intensity for different regions of the input images. As we can see, for a sharp bordered lesion, the classification model mainly focuses on the lesion itself (which is clinically relevant as lesion shapes, colors etc. can be good indicators for the lesion being cancerous or not) and less on the surrounding skin, however for a blurry bordered lesion it tends to rather find the surrounding skin relevant and not the lesion itself, so in this case the model focuses mainly on irrelevant background regions. These insights are extremely important for model builders & developers when evaluating their model’s robustness and reliability, as well as to non-experts (clinicians, doctors, or laypersons who are the users of different skin cancer detection apps) for interpretability and calibrating proper trust in the skin cancer classification model.
3.1.3. LIME for the classification model¶
In [26]:
torch.cuda.empty_cache()
# Define LIME explainer
explainer = lime_image.LimeImageExplainer()
# Predict function for LIME
def predict_fn(images):
images = [transform(transforms.ToPILImage()(img)) for img in images] # Transform on CPU
images = torch.stack(images).to(device) # Then move to GPU
with torch.no_grad():
outputs = model_class(images)
torch.cuda.empty_cache() # Clear unused memory
return outputs.cpu().detach().numpy()
# Test LIME on a single image
image, label = test_dataset.__getitem__(0)
image_np = np.array(transforms.ToPILImage()(image))
# Generate explanation
explanation = explainer.explain_instance(
image_np,
predict_fn,
labels=[label],
hide_color=0,
num_samples=1000
)
# Get the explanation for the specific class label
temp, mask = explanation.get_image_and_mask(
label,
positive_only=True, # Only show positive contributions
num_features=3, # Number of features (segments) to show
hide_rest=False
)
# Visualize the explanation using mark_boundaries to overlay the mask on the image
fig, ax = plt.subplots(figsize=(8, 8))
ax.imshow(mark_boundaries(temp, mask))
# Title
ax.set_title(f'LIME Explanation for Class {label}')
ax.axis('off')
# Create a colorbar to show positive and negative contributions
cmap = plt.cm.RdYlGn # Red-Yellow-Green color map, with red for negative and green for positive
# norm = mpl.colors.Normalize(vmin=0, vmax=1) # Normalize the color scale to [0, 1]
# Add colorbar
sm = plt.cm.ScalarMappable(cmap=cmap)
sm.set_array([]) # Create a dummy array for the color bar
cbar = plt.colorbar(sm, ax=ax, fraction=0.046, pad=0.04)
cbar.set_label('Feature Contribution')
# Display the plot
plt.show()
0%| | 0/1000 [00:00<?, ?it/s]
This LIME plot with hide_rest=False parameter setting shows that the trained model's main focus is not on the lesion's border, but rather on some parts of the lesion itself, for example some internal pigmentation patterns (just as we can see on the first Grad-Cam plot above, for sharp bordered lesion), as well as on the surrounding skin (just as we can see on the second Grad-Cam plot above, for blurry bordered lesion). So the influential regions that contributed to the model's prediction are mainly those regions that are parts of the lesion (but not necessarily border areas) and/or those that are parts of the surrounding skin. This can be mainly useful for model builders and developers when wanting to visualise what the model thinks as most influential areas when clasifying this specific instance. It can also be interesting from a non-experts perspective (for example a doctor or clinician), as they could compare the influential areas marked by the model with dermatological best-practices to look at when examining a lesion.
(The purple colors on this image come from the image transformation that is needed for the classification model, and not a result of the LIME XAI approach.)
In [27]:
# Define LIME explainer
explainer = lime_image.LimeImageExplainer()
# Test LIME on a single image
image, label = test_dataset.__getitem__(122)
image_np = np.array(transforms.ToPILImage()(image))
# Generate explanation
explanation = explainer.explain_instance(
image_np,
predict_fn,
labels=[label],
hide_color=0,
num_samples=300
)
# Get the explanation for the specific class label
temp, mask = explanation.get_image_and_mask(
label,
positive_only=True, # Only show positive contributions
num_features=3, # Number of features (segments) to show
hide_rest=False
)
# Visualize the explanation using mark_boundaries to overlay the mask on the image
fig, ax = plt.subplots(figsize=(8, 8))
ax.imshow(mark_boundaries(temp, mask))
# Title
ax.set_title(f'LIME Explanation for Class {label}')
ax.axis('off')
# Create a colorbar to show positive and negative contributions
cmap = plt.cm.RdYlGn # Red-Yellow-Green color map, with red for negative and green for positive
# norm = mpl.colors.Normalize(vmin=0, vmax=1) # Normalize the color scale to [0, 1]
# Add colorbar
sm = plt.cm.ScalarMappable(cmap=cmap)
sm.set_array([]) # Create a dummy array for the color bar
cbar = plt.colorbar(sm, ax=ax, fraction=0.046, pad=0.04)
cbar.set_label('Feature Contribution')
# Display the plot
plt.show()
0%| | 0/300 [00:00<?, ?it/s]
This second LIME plot with hide_rest=False parameter setting also underlines our previous findings that the trained model's main focus is not on the lesion's border, but rather on some parts of the lesion itself, for example some internal pigmentation patterns. On this specific instance, however we can see that less area is marked as influential from the surrounding skin area, but still some of the skin areas that are relatively far away from the lesion are marked as influential for the models classification. This could be interesting for model builders and developers to investigate further (for example to find out whether it is common for correctly or incorrectly classified images to focus on the surrounding skin.)
(Again, the purple colors on this image come from the image transformation that is needed for the classification model, and not a result of the LIME XAI approach.)
3.1.4 Entropy visualization¶
In [28]:
multiclass_categories = {
0: 'NV',
1: 'MEL',
2: 'BCC',
3: 'AKIEC',
4: 'DF',
5: 'VASC'
}
class MobileNetV3(nn.Module):
def __init__(self, num_classes):
super(MobileNetV3, self).__init__()
self.model = models.mobilenet_v3_small(pretrained=True)
in_features = self.model.classifier[3].in_features
self.model.classifier[3] = nn.Linear(in_features, num_classes)
self.softmax = nn.Softmax(dim=1)
def forward(self, x):
x = self.model(x)
x = self.softmax(x)
return x
model = MobileNetV3(num_classes=6)
model.load_state_dict(torch.load("/content/drive/MyDrive/XAI/best_model_classification.pth", map_location=torch.device('cpu')))
model.eval()
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.
warnings.warn(
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=MobileNet_V3_Small_Weights.IMAGENET1K_V1`. You can also use `weights=MobileNet_V3_Small_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
<ipython-input-28-52c42faf06a3>:25: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
model.load_state_dict(torch.load("/content/drive/MyDrive/XAI/best_model_classification.pth", map_location=torch.device('cpu')))
Out[28]:
MobileNetV3(
(model): MobileNetV3(
(features): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(16, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=16, bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(16, 8, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(8, 16, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(2): Conv2dNormActivation(
(0): Conv2d(16, 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(2): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(16, 72, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(72, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Conv2dNormActivation(
(0): Conv2d(72, 72, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=72, bias=False)
(1): BatchNorm2d(72, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(2): Conv2dNormActivation(
(0): Conv2d(72, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(24, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(3): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(24, 88, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(88, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(1): Conv2dNormActivation(
(0): Conv2d(88, 88, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=88, bias=False)
(1): BatchNorm2d(88, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
)
(2): Conv2dNormActivation(
(0): Conv2d(88, 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(24, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(4): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(24, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96, bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(96, 24, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(24, 96, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(96, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(5): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(240, 64, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(64, 240, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(6): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
(1): BatchNorm2d(240, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(240, 64, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(64, 240, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(40, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(7): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(40, 120, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(120, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
(1): BatchNorm2d(120, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(120, 32, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(32, 120, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(120, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(48, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(8): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(48, 144, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(144, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(144, 144, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=144, bias=False)
(1): BatchNorm2d(144, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(144, 40, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(40, 144, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(144, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(48, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(9): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(48, 288, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(288, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(288, 288, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=288, bias=False)
(1): BatchNorm2d(288, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(288, 72, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(72, 288, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(288, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(10): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(576, 576, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=576, bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(576, 144, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(144, 576, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(576, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(11): InvertedResidual(
(block): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(1): Conv2dNormActivation(
(0): Conv2d(576, 576, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=576, bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
(2): SqueezeExcitation(
(avgpool): AdaptiveAvgPool2d(output_size=1)
(fc1): Conv2d(576, 144, kernel_size=(1, 1), stride=(1, 1))
(fc2): Conv2d(144, 576, kernel_size=(1, 1), stride=(1, 1))
(activation): ReLU()
(scale_activation): Hardsigmoid()
)
(3): Conv2dNormActivation(
(0): Conv2d(576, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
)
)
)
(12): Conv2dNormActivation(
(0): Conv2d(96, 576, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(576, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
(2): Hardswish()
)
)
(avgpool): AdaptiveAvgPool2d(output_size=1)
(classifier): Sequential(
(0): Linear(in_features=576, out_features=1024, bias=True)
(1): Hardswish()
(2): Dropout(p=0.2, inplace=True)
(3): Linear(in_features=1024, out_features=6, bias=True)
)
)
(softmax): Softmax(dim=1)
)
In [29]:
def calculate_entropy(probs):
"""Compute entropy for a probability distribution."""
return -torch.sum(probs * torch.log(probs + 1e-8), dim=1) # Add epsilon for numerical stability
def preprocess_image(image_path):
"""Preprocess image for MobileNetV3."""
transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) # Normalization for ImageNet
])
image = Image.open(image_path).convert('RGB')
return transform(image).unsqueeze(0) # Add batch dimension
folder_path = "/content/extracted_folder/images"
image_extensions = {'.jpg'}
labels = ['NV', 'MEL', 'BCC', 'AKIEC', 'DF', 'VASC']
NV_ent = []
MEL_ent = []
BCC_ent = []
AKIEC_ent = []
DF_ent = []
VASC_ent = []
def predict_with_uncert(model, image_path):
"""Get class probabilities and uncertainty scores for an image."""
image_tensor = preprocess_image(image_path)
with torch.no_grad():
probs = model(image_tensor)
entropy = calculate_entropy(probs)
labels = ['NV', 'MEL', 'BCC', 'AKIEC', 'DF', 'VASC']
return probs, entropy
for file_name in os.listdir(folder_path):
file_path = os.path.join(folder_path, file_name)
#print(file_path)
# Check if the file is an image
if os.path.isfile(file_path) and os.path.splitext(file_name)[1].lower() in image_extensions:
# Call the function for the image
probs, entropy = predict_with_uncert(model, file_path)
max_index = torch.argmax(probs, dim=1).item()
#print(probs)
#print(max_index)
# Get the label corresponding to the highest value
max_label = labels[max_index]
if max_label == labels[0]:
NV_ent.append(entropy.item())
if max_label == labels[1]:
MEL_ent.append(entropy.item())
if max_label == labels[2]:
BCC_ent.append(entropy.item())
if max_label == labels[3]:
AKIEC_ent.append(entropy.item())
if max_label == labels[4]:
DF_ent.append(entropy.item())
if max_label == labels[5]:
VASC_ent.append(entropy.item())
NV_entropy = np.mean(NV_ent)
MEL_entropy = np.mean(MEL_ent)
BCC_entropy = np.mean(BCC_ent)
AKIEC_entropy = np.mean(AKIEC_ent)
DF_entropy = np.mean(DF_ent)
VASC_entropy = np.mean(VASC_ent)
print(f"NV Entropy:{NV_entropy}, Nr. of classified images: {len(NV_ent)}")
print(f"MEL Entropy:{MEL_entropy}, Nr. of classified images: {len(MEL_ent)}")
print(f"BCC Entropy:{BCC_entropy}, Nr. of classified images: {len(BCC_ent)}")
print(f"AKIEC Entropy:{AKIEC_entropy}, Nr. of classified images: {len(AKIEC_ent)}")
print(f"DF Entropy:{DF_entropy}, Nr. of classified images: {len(DF_ent)}")
print(f"VASC Entropy:{VASC_entropy}, Nr. of classified images: {len(VASC_ent)}")
import matplotlib.pyplot as plt
entropies = {
'NV': NV_entropy,
'MEL': MEL_entropy,
'BCC': BCC_entropy,
'AKIEC': AKIEC_entropy,
'DF': DF_entropy,
'VASC': VASC_entropy
}
# Extract labels and values for plotting
labels = list(entropies.keys())
values = list(entropies.values())
# Create the bar plot
plt.figure(figsize=(10, 6))
plt.bar(labels, values)
plt.xlabel("Classes")
plt.ylabel("Average Entropy")
plt.title("Average Entropy per Class")
plt.xticks(rotation=45, ha='right') # Rotate x-axis labels
plt.tight_layout()
plt.show()
NV Entropy:0.12493987881531567, Nr. of classified images: 5419 MEL Entropy:0.23018173988135066, Nr. of classified images: 2340 BCC Entropy:0.29259641102033584, Nr. of classified images: 551 AKIEC Entropy:0.3047801893422042, Nr. of classified images: 656 DF Entropy:0.4340517743669074, Nr. of classified images: 194 VASC Entropy:0.17189825866071967, Nr. of classified images: 855
In this plot, the average entropy for each class is shown, indicating how certain the model is about the classifications for the individual classes. For the class “melanocytic nevi” (NV), for example, the average entropy is lower than for all other classes. This indicates that predictions for this class have the lowest uncertainty. The class “dermatofibroma” (DF), in contrast, has a higher average entropy. Consequently, the model is rather unsure about the DF classifications. Model developers could especially focus on this class to decrease the uncertainty of the model.
3.2 XAI approaches for the segmentation model¶
3.2.1 Grad-Cam for different layers of the network¶
In [30]:
# Ensure Keras backend is TensorFlow
os.environ["KERAS_BACKEND"] = "tensorflow"
# Use your fuzzy net segmentation model
model_path = "/content/drive/MyDrive/XAI/Skin_cancer_Segmentation.keras" # Update with your model file path
model_seg = load_model(model_path, custom_objects={"dice_coefficient": dice_coefficient, "iou": iou})
# Update the last convolutional layer name for your fuzzy net model
#last_conv_layer_name = "conv2d_23" # Update based on your model architecture
# Load and preprocess the input image
def preprocess_input(img_path, target_size=(256, 256)):
"""Load and preprocess an image for the model."""
img = tf.keras.utils.load_img(img_path, target_size=target_size)
img_array = tf.keras.utils.img_to_array(img)
img_array = img_array / 255.0 # Normalize to [0, 1]
return np.expand_dims(img_array, axis=0)
# Grad-CAM heatmap generation
def make_gradcam_heatmap(img_array, model, last_conv_layer_name, target_class=None):
"""Generate a Grad-CAM heatmap for a segmentation model."""
# Create a model mapping input to last conv layer and output
grad_model = keras.models.Model(
[model.inputs],
[model.get_layer(last_conv_layer_name).output, model.output]
)
# Compute gradients for the target class
with tf.GradientTape() as tape:
last_conv_layer_output, predictions = grad_model(img_array)
# Use the first class if none is provided
if target_class is None:
target_class = 0 # For binary segmentation
# Focus on the target class across the entire output
class_channel = predictions[..., target_class]
# Compute gradients of the class channel wrt the feature map
grads = tape.gradient(class_channel, last_conv_layer_output)
# Pool the gradients to get global importance of each channel
pooled_grads = tf.reduce_mean(grads, axis=(0, 1, 2))
# Weight the feature map by importance
last_conv_layer_output = last_conv_layer_output[0]
heatmap = last_conv_layer_output @ pooled_grads[..., tf.newaxis]
heatmap = tf.squeeze(heatmap)
# Normalize heatmap to [0, 1]
heatmap = tf.maximum(heatmap, 0) / tf.math.reduce_max(heatmap)
return heatmap.numpy()
# Visualize the Grad-CAM heatmap
def overlay_heatmap(heatmap, img, alpha=0.5, colormap="jet"):
"""Overlay the Grad-CAM heatmap on the original image."""
heatmap = np.uint8(255 * heatmap) # Scale to [0, 255]
colormap = plt.get_cmap(colormap)
colored_heatmap = colormap(heatmap)
colored_heatmap = np.delete(colored_heatmap, 3, axis=2) # Remove alpha channel
colored_heatmap = tf.image.resize(colored_heatmap, (img.shape[0], img.shape[1])).numpy()
superimposed_img = colored_heatmap * alpha + img / 255.0 # Normalize original image
superimposed_img = np.clip(superimposed_img, 0, 1)
return (superimposed_img * 255).astype("uint8")
# List of layers to compare
layer_names = [
"conv2d_3", # Early layer
"conv2d_7", # Mid layer
"conv2d_10", # Mid layer
"conv2d_23" # Final layer
]
# Function to generate and plot Grad-CAM for multiple layers
def plot_gradcam_for_layers(model, img_array, layer_names, target_class=0):
"""Plot Grad-CAM heatmaps for a list of layers."""
num_layers = len(layer_names)
fig, axes = plt.subplots(1, num_layers, figsize=(20, 5))
for i, layer_name in enumerate(layer_names):
# Generate Grad-CAM heatmap for the layer
heatmap = make_gradcam_heatmap(img_array, model, layer_name, target_class=target_class)
# Plot heatmap
axes[i].imshow(heatmap, cmap="jet")
axes[i].set_title(f"Layer: {layer_name}")
axes[i].axis("off")
plt.tight_layout()
plt.show()
# Preprocess the input image
img_path = "/content/extracted_folder/images/ISIC_0024317.jpg" # Path to your image
img_array = preprocess_input(img_path, target_size=(256, 256))
# Call the function
plot_gradcam_for_layers(model_seg, img_array, layer_names, target_class=0)
In [31]:
# List of layers to compare
layer_names = [
"batch_normalization_3", # Early layer
"batch_normalization_10", # Mid layer
"batch_normalization_14", # Mid layer
"batch_normalization_18" # Final layer
]
# Function to generate and plot Grad-CAM for multiple layers
def plot_gradcam_for_layers(model, img_array, layer_names, target_class=0):
"""Plot Grad-CAM heatmaps for a list of layers."""
num_layers = len(layer_names)
fig, axes = plt.subplots(1, num_layers, figsize=(20, 5))
for i, layer_name in enumerate(layer_names):
# Generate Grad-CAM heatmap for the layer
heatmap = make_gradcam_heatmap(img_array, model, layer_name, target_class=target_class)
# Plot heatmap
axes[i].imshow(heatmap, cmap="jet")
axes[i].set_title(f"Layer: {layer_name}")
axes[i].axis("off")
plt.tight_layout()
plt.show()
# Preprocess the input image
img_path = "/content/extracted_folder/images/ISIC_0024317.jpg"
img_array = preprocess_input(img_path, target_size=(256, 256))
# Call the function
plot_gradcam_for_layers(model_seg, img_array, layer_names, target_class=0)
Conclusion
The Grad-CAM visualizations across the Fuzzy U-Net layers reveal a clear progression of feature learning. Early layers focus on low-level features such as texture and edges, while mid-layers refine attention toward lesion-specific patterns like pigmentation and borders. Late layers exhibit precise boundary detection, aligning closely with the lesion's ground truth.
These insights are critical for:
- Model Debugging: Ensuring the network learns relevant features at different stages.
- Trust Building: Demonstrating that the model focuses on clinically meaningful regions, enhancing confidence in predictions.
- Performance Validation: Comparing the segmentation quality across layers to refine architecture or training methods.
Grad-CAM provides a powerful interpretability tool for segmentation models, aiding machine learning practitioners, clinicians, and researchers in understanding and improving model behavior.
3.2.2 Occlusion sensitivity¶
In [32]:
# Parameters for occlusion
occluding_size = 30 # Size of the occluding patch
occluding_pixel = (0, 0, 0) # Pixel value for occlusion (black in RGB)
occluding_stride = 10 # Stride for occlusion
def preprocess_image(img_path, target_size=(256, 256)):
"""
Preprocess the input image for the model using Pillow.
"""
image = Image.open(img_path).resize(target_size)
image = np.array(image) / 255.0 # Normalize to [0, 1]
return image
def load_mask(mask_path, target_size=(256, 256)):
"""
Load the segmentation mask from the given path.
"""
mask = Image.open(mask_path).resize(target_size)
mask = np.array(mask) / 255.0 # Normalize mask values to [0, 1]
return mask
def occlusion_sensitivity(image, model, target_class=0, occluding_size=50, occluding_pixel=(0, 0, 0), occluding_stride=10):
"""
Perform occlusion sensitivity analysis.
Args:
image: Input image as a NumPy array.
model: Pre-trained model.
target_class: Target class index (default: 0 for binary).
occluding_size: Size of the occlusion patch.
occluding_pixel: RGB tuple for occlusion (default: black).
occluding_stride: Stride of occlusion.
Returns:
heatmap: Occlusion sensitivity heatmap.
occluded_example: An example of the occluded image at the final position.
"""
# Get image dimensions
height, width, _ = image.shape
# Calculate heatmap dimensions
output_height = math.ceil((height - occluding_size) / occluding_stride + 1)
output_width = math.ceil((width - occluding_size) / occluding_stride + 1)
heatmap = np.zeros((int(output_height), int(output_width)))
# Create an occluded example for visualization
occluded_example = None
for h in range(0, int(output_height)):
for w in range(0, int(output_width)):
# Calculate occlusion region
h_start = h * occluding_stride
w_start = w * occluding_stride
h_end = min(height, h_start + occluding_size)
w_end = min(width, w_start + occluding_size)
# Create a copy of the image and apply the occlusion
occluded_image = copy.deepcopy(image)
occluded_image[h_start:h_end, w_start:w_end, :] = occluding_pixel
# Preprocess the occluded image and predict
input_image = np.expand_dims(occluded_image, axis=0) # Add batch dimension
predictions = model.predict(input_image)
# Record the probability for the target class
prob = predictions[0][..., target_class].mean() # Adjust for segmentation
heatmap[h, w] = prob
# Save one example occluded image for visualization
if h == int(output_height / 2) and w == int(output_width / 2):
occluded_example = occluded_image
return heatmap, occluded_example
# Path to the input image and mask
img_path = "/content/extracted_folder/images/ISIC_0024810.jpg"
mask_path = "/content/extracted_folder/masks/ISIC_0024810_segmentation.png"
# Preprocess the image and mask
original_img = preprocess_image(img_path, target_size=(256, 256))
segmentation_mask = load_mask(mask_path, target_size=(256, 256))
# Generate occlusion sensitivity heatmap
sensitivity_heatmap, occluded_example = occlusion_sensitivity(
original_img,
model_seg,
target_class=0, # For binary segmentation
occluding_size=occluding_size,
occluding_pixel=occluding_pixel,
occluding_stride=occluding_stride
)
# Plot the results with improved layout and resizing
plt.figure(figsize=(16, 8)) # Adjust the figure size for better visualization
# Original Image
plt.subplot(1, 4, 1)
plt.imshow(original_img)
plt.title("Original Image", fontsize=14)
plt.axis("off")
# Occluded Image
plt.subplot(1, 4, 2)
plt.imshow(occluded_example)
plt.title("Occluded Image", fontsize=14)
plt.axis("off")
# Sensitivity Heatmap
plt.subplot(1, 4, 3)
sns.heatmap(sensitivity_heatmap, cmap="viridis", cbar=True, xticklabels=False, yticklabels=False, square=True)
plt.title("Occlusion Sensitivity Heatmap", fontsize=14)
plt.axis("off")
# Segmentation Mask
plt.subplot(1, 4, 4)
plt.imshow(segmentation_mask, cmap="gray")
plt.title("Segmentation Mask", fontsize=14)
plt.axis("off")
# Tighten layout for better spacing
plt.tight_layout()
plt.show()
1/1 ━━━━━━━━━━━━━━━━━━━━ 5s 5s/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 43ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 45ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 42ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 39ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 45ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 44ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 43ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 48ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step
In [33]:
# Parameters for occlusion
occluding_size = 30 # Size of the occluding patch
occluding_pixel = (0, 0, 0) # Pixel value for occlusion (black in RGB)
occluding_stride = 10 # Stride for occlusion
def preprocess_image(img_path, target_size=(256, 256)):
"""
Preprocess the input image for the model using Pillow.
"""
image = Image.open(img_path).resize(target_size)
image = np.array(image) / 255.0 # Normalize to [0, 1]
return image
def load_mask(mask_path, target_size=(256, 256)):
"""
Load the segmentation mask from the given path.
"""
mask = Image.open(mask_path).resize(target_size)
mask = np.array(mask) / 255.0 # Normalize mask values to [0, 1]
return mask
def occlusion_sensitivity(image, model, target_class=0, occluding_size=50, occluding_pixel=(0, 0, 0), occluding_stride=10):
"""
Perform occlusion sensitivity analysis.
Args:
image: Input image as a NumPy array.
model: Pre-trained model.
target_class: Target class index (default: 0 for binary).
occluding_size: Size of the occlusion patch.
occluding_pixel: RGB tuple for occlusion (default: black).
occluding_stride: Stride of occlusion.
Returns:
heatmap: Occlusion sensitivity heatmap.
occluded_example: An example of the occluded image at the final position.
"""
# Get image dimensions
height, width, _ = image.shape
# Calculate heatmap dimensions
output_height = math.ceil((height - occluding_size) / occluding_stride + 1)
output_width = math.ceil((width - occluding_size) / occluding_stride + 1)
heatmap = np.zeros((int(output_height), int(output_width)))
# Create an occluded example for visualization
occluded_example = None
for h in range(0, int(output_height)):
for w in range(0, int(output_width)):
# Calculate occlusion region
h_start = h * occluding_stride
w_start = w * occluding_stride
h_end = min(height, h_start + occluding_size)
w_end = min(width, w_start + occluding_size)
# Create a copy of the image and apply the occlusion
occluded_image = copy.deepcopy(image)
occluded_image[h_start:h_end, w_start:w_end, :] = occluding_pixel
# Preprocess the occluded image and predict
input_image = np.expand_dims(occluded_image, axis=0) # Add batch dimension
predictions = model.predict(input_image)
# Record the probability for the target class
prob = predictions[0][..., target_class].mean() # Adjust for segmentation
heatmap[h, w] = prob
# Save one example occluded image for visualization
if h == int(output_height / 2) and w == int(output_width / 2):
occluded_example = occluded_image
return heatmap, occluded_example
# Path to the input image and mask
img_path = "/content/extracted_folder/images/ISIC_0024448.jpg"
mask_path = "/content/extracted_folder/masks/ISIC_0024448_segmentation.png"
# Preprocess the image and mask
original_img = preprocess_image(img_path, target_size=(256, 256))
segmentation_mask = load_mask(mask_path, target_size=(256, 256))
# Generate occlusion sensitivity heatmap
sensitivity_heatmap, occluded_example = occlusion_sensitivity(
original_img,
model_seg,
target_class=0, # For binary segmentation
occluding_size=occluding_size,
occluding_pixel=occluding_pixel,
occluding_stride=occluding_stride
)
# Plot the results with improved layout and resizing
plt.figure(figsize=(16, 8)) # Adjust the figure size for better visualization
# Original Image
plt.subplot(1, 4, 1)
plt.imshow(original_img)
plt.title("Original Image", fontsize=14)
plt.axis("off")
# Occluded Image
plt.subplot(1, 4, 2)
plt.imshow(occluded_example)
plt.title("Occluded Image", fontsize=14)
plt.axis("off")
# Sensitivity Heatmap
plt.subplot(1, 4, 3)
sns.heatmap(sensitivity_heatmap, cmap="viridis", cbar=True, xticklabels=False, yticklabels=False, square=True)
plt.title("Occlusion Sensitivity Heatmap", fontsize=14)
plt.axis("off")
# Segmentation Mask
plt.subplot(1, 4, 4)
plt.imshow(segmentation_mask, cmap="gray")
plt.title("Segmentation Mask", fontsize=14)
plt.axis("off")
# Tighten layout for better spacing
plt.tight_layout()
plt.show()
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 50ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 47ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 52ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 42ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 17ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 59ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 41ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 48ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 45ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 46ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 44ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 49ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 78ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 85ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 75ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 92ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 80ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 182ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 103ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 72ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 104ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 107ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 206ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 112ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 112ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 47ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 81ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 76ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 71ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 90ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 59ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 69ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 49ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 73ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 64ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 75ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 55ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 55ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 52ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 46ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 39ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 102ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 46ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 39ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 44ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 66ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 42ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 87ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 51ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 52ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 91ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 121ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 56ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 71ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 61ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 146ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 172ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 191ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 156ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 49ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step
In [34]:
# Parameters for occlusion
occluding_size = 30 # Size of the occluding patch
occluding_pixel = (0, 0, 0) # Pixel value for occlusion (black in RGB)
occluding_stride = 10 # Stride for occlusion
# Path to the input image and mask
img_path = "/content/extracted_folder/images/ISIC_0024354.jpg"
mask_path = "/content/extracted_folder/masks/ISIC_0024354_segmentation.png"
# Preprocess the image and mask
original_img = preprocess_image(img_path, target_size=(256, 256))
segmentation_mask = load_mask(mask_path, target_size=(256, 256))
# Generate occlusion sensitivity heatmap
sensitivity_heatmap, occluded_example = occlusion_sensitivity(
original_img,
model_seg,
target_class=0, # For binary segmentation
occluding_size=occluding_size,
occluding_pixel=occluding_pixel,
occluding_stride=occluding_stride
)
# Plot the results with improved layout and resizing
plt.figure(figsize=(16, 8)) # Adjust the figure size for better visualization
# Original Image
plt.subplot(1, 4, 1)
plt.imshow(original_img)
plt.title("Original Image", fontsize=14)
plt.axis("off")
# Occluded Image
plt.subplot(1, 4, 2)
plt.imshow(occluded_example)
plt.title("Occluded Image", fontsize=14)
plt.axis("off")
# Sensitivity Heatmap
plt.subplot(1, 4, 3)
sns.heatmap(sensitivity_heatmap, cmap="viridis", cbar=True, xticklabels=False, yticklabels=False, square=True)
plt.title("Occlusion Sensitivity Heatmap", fontsize=14)
plt.axis("off")
# Segmentation Mask
plt.subplot(1, 4, 4)
plt.imshow(segmentation_mask, cmap="gray")
plt.title("Segmentation Mask", fontsize=14)
plt.axis("off")
# Tighten layout for better spacing
plt.tight_layout()
plt.show()
1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 41ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 45ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 48ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 42ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 45ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 39ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 41ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 36ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 24ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 18ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 33ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 21ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 25ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 20ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 19ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 23ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 22ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 37ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 26ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 38ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 28ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 31ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 27ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 30ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 40ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 34ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 29ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 35ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 32ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 39ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 41ms/step 1/1 ━━━━━━━━━━━━━━━━━━━━ 0s 43ms/step
Conclusion¶
In the heatmaps:
- Yellow regions indicate areas where occlusion has a significant impact on the model’s segmentation performance.
- Blue regions indicate areas where occlusion has little to no effect.
First example:
- The lesion is clearly visible, with irregular pigmentation and boundaries.
- The occlusion sensitivity heatmap strongly highlights the lesion boundaries, indicating that the model is paying significant attention to these region for its predictions.
Second and Third example:
- The lesion is more subtle, with diffuse pigmentation and less clear boundaries.
- The heatmap shows less focused and more dispersed attention, with less solid and sharp lesion borders.
- This could indicate that the model is uncertain about the lesion's boundaries, possibly due to: Vairablility in the lesion's texture and color.
- The mask aligns with the sensitive central area, though the lack of sensitivity at the edges may indicate potential weaknesses in boundary detection.
Who can benefit from these plots?
Clinicians: These heatmaps validate that the model focuses on clinically relevant areas, such as the lesion core and boundary. This builds trust in the model's decisions.
Researchers: Insights into which parts of the lesion are most critical for predictions can guide the development of more robust models.
Machine Learning Practitioners: Occlusion sensitivity highlights potential weaknesses in the model (e.g., over-reliance on the lesion core), suggesting areas for refinement in training or architecture.